SNHG5 Promotes Colorectal Cancer Cell Survival by Counteracting STAU1-Mediated Mrna Destabilization

ARTICLE

Received 2 Aug 2016 | Accepted 8 Nov 2016 | Published 22 Dec 2016 DOI: 10.1038/ncomms13875 OPEN SNHG5 promotes colorectal cancer cell survival by counteracting STAU1-mediated mRNA destabilization

Nkerorema Djodji Damas1, Michela Marcatti1, Christophe Coˆme1, Lise Lotte Christensen2, Morten Muhlig Nielsen2, Roland Baumgartner1, Helene Maria Gylling1, Giulia Maglieri1, Carsten Friis Rundsten1, Stefan E. Seemann3, Nicolas Rapin1, Simon The´zenas4, Søren Vang2, Torben Ørntoft2, Claus Lindbjerg Andersen2, Jakob Skou Pedersen2 & Anders H. Lund1

We currently have limited knowledge of the involvement of long non-coding RNAs (lncRNAs) in normal cellular processes and pathologies. Here, we identify and characterize SNHG5 as a stable cytoplasmic lncRNA with up-regulated expression in colorectal cancer. Depletion of SNHG5 induces cell cycle arrest and apoptosis in vitro and limits tumour outgrowth in vivo, whereas SNHG5 overexpression counteracts oxaliplatin-induced apoptosis. Using an unbiased approach, we identify 121 transcript sites interacting with SNHG5 in the cytoplasm. Impor- tantly, knockdown of key SNHG5 target transcripts, including SPATS2, induces apoptosis and thus mimics the effect seen following SNHG5 depletion. Mechanistically, we suggest that SNHG5 stabilizes the target transcripts by blocking their degradation by STAU1. Accordingly, depletion of STAU1 rescues the apoptosis induced after SNHG5 knockdown. Hence, we characterize SNHG5 as a lncRNA promoting tumour cell survival in colorectal cancer and delineate a novel mechanism in which a cytoplasmic lncRNA functions through blocking the action of STAU1.

1 Biotech Research and Innovation Centre, University of Copenhagen, Ole Maaloes Vej 5 2200 Copenhagen, Denmark. 2 Department for Molecular Medicine, Aarhus University Hospital, Brendstrupgaardsvej 21 8000 Aarhus, Denmark. 3 Center for Non-Coding RNA in Technology and Health, University of Copenhagen, Grønnegårdsvej 3 1870, Frederiksberg, Denmark. 4 Biostatistics Unit, Institut Re´gional du Cancer de Montpellier (ICM)—Val d’Aurelle, 208 Avenue des Apothicaires, 34298 Montpellier, France. Correspondence and requests for materials should be addressed to A.H.L. (email [email protected]).

NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875

berrant proliferation and increased cell survival are key Results processes during malignant transformation and tumour SNHG5 is up-regulated in colorectal cancer. To identify Aprogression. To facilitate these processes, a number non-coding RNAs deregulated in CRC, we profiled their of chromosomal rearrangements, mutations and epigenetic expression in a cohort of 44 carcinomas, 39 adenomas, 20 adja- modifications are typically selected for in cancerous cells, cent normal mucosa and 10 CRC cell lines using a previously resulting in an overall alteration of gene expression. Colorectal described custom-designed microarray platform18. As expected, carcinoma (CRC) is the third most common cancer worldwide1. the overall expression level of lncRNAs was lower than that of the While invasive colorectal cancers that have not yet compro- protein-coding genes in all sample sets (Supplementary Fig. 1a)19. mised regional lymph nodes (stage I–II) have relatively good Among the most significantly deregulated transcripts, we prognoses with the current treatments and are curable in 73% of identified several lncRNAs previously found deregulated in the cases, the progression of the disease is fast and untreated cancer, such as ncRAN and GAS5, indicating the validity of our tumours rapidly disseminate to lymph nodes (stage III) and approach (Supplementary Fig. 1b,c). We focus here on SNHG5 metastasize to distant sites (stage IV)2. Thus, a better for which no function in CRC has been ascribed so far. RNA-seq understanding of the mechanisms driving the disease and confirmed the up-regulation of SNHG5 in CRC in a larger identification of additional therapeutic targets is a priority for independent cohort of 294 normal colon mucosa samples, improving CRC treatment3. Several studies have pointed to the 33 adenomas and 280 CRC samples (n ¼ 313) (Fig. 1a). This emerging roles of long non-coding RNAs (lncRNAs) in tumour was confirmed also restricting the analysis to the 268 paired development, which could provide new candidates for diagnostics samples included in the same cohort (Supplementary Fig. 2a). and therapy. Interestingly, we find SNHG5 to be significantly up-regulated Although mammalian genomes are widely transcribed, only both between normal tissues and adenomas, and from adenomas 1–2% of the genomic output encodes for proteins4,5. Among to carcinoma stage I (malignant tumour), suggesting SNHG5 the large fraction of non-coding transcripts, the class of lncRNAs, up-regulation as an early event in CRC development (Fig. 1b). arbitrarily defined as transcripts longer than 200 nts, is receiving SNHG5 is composed of six exons spliced into two main increasing attention and may present new opportunities isoforms: the 1032 nts long SNHG5-1 and the shorter SNHG5-2 for cancer diagnosis and treatment. Although thousands of (430 nts) (Fig. 1c). SNHG5-2 is the most abundant isoform lncRNAs have been identified, we still lack insight into the expressed in CRC (Supplementary Fig. 2b), and we refer to it structural–functional significance of the vast majority of these as SNHG5 in the following experiments. Two snoRNAs, U50A molecules in regulating fundamental cellular processes. However, and U50B, are encoded in introns 4 and 5, respectively. We extensive gene expression and copy number variation analyses profiled the expression of SNHG5 in a panel of human tissues and have linked alteration of lncRNA expression to tumour found it most highly expressed in muscles, liver and development. Resultantly, several lncRNAs, such as MALAT1, heart (Supplementary Fig. 2c). Reverse transcription (RT)–PCR HOTAIR, ANRIL, PVT1 and lincRNA-p21, have been reported analysis revealed that the SNHG5 transcript, like many to play significant roles in cancer initiation and development6–10. other mRNA-like non-coding RNAs, is poly-adenylated In this study, we aimed to identify and characterize lncRNAs (Supplementary Fig. 2d). The SNHG5 transcript has a half-life functionally impacting on CRC. Profiling a large set of CRCs, (t½) of 1 h, indicating an intermediate turnover rate when we identified the cytoplasmic lncRNA SNHG5 as signifi- compared with transcripts with a very short half-life, such cantly overexpressed in tumours. We provide evidence that as MYC (t½ ¼ 30 min), or more stable mRNAs, such as SNHG5 expression regulates the survival of CRC cells and the ACTB (t½ ¼ 5–6 h) (Supplementary Fig. 2e). RNA fluorescence progression of CRC tumour xenografts in a mouse model. in situ hybridization (RNA-FISH) was performed using LNA Importantly, whereas knockdown of SNHG5 prominently induces probes spanning the exon–exon junctions 3–4 and 4–5, apoptosis, SNHG5 overexpression can protect CRC cells from respectively, and showed SNHG5 to be predominantly cytosolic oxaliplatin-induced apoptosis. While most of the hitherto in both HCT116 CRC cells and in immortalized TIG-3 human characterized lncRNAs function in the nucleus, much less is fibroblasts (Fig. 1d). This was also demonstrated by subcellular known about lncRNAs and their mode of action in the fractionation studies and qRT–PCR (Fig. 1e). Polysome fractio- cytoplasmic compartment. A notable exception being competing nation studies showed a clear SNHG5 enrichment in the light, un- endogenous RNAs (ceRNAs), which act as molecular sponges for translated fraction suggesting no association with ribosomes microRNAs hence relieving repression of target mRNAs11,12. (Supplementary Fig. 2f). In conclusion, we characterize SNHG5 to Other known mechanisms for lncRNAs in the cytoplasm invo- be a stable and poly-adenylated ncRNA transcript not associated lve post-transcriptional regulation affecting mRNA stability or with ribosomes and with an up-regulated expression in CRC. accessibility to the translational machinery13. Through an unbiased forward identification of mRNAs interacting with SNHG5 in the cytoplasm, we identify 121 interacting transcript sites in SNHG5 affects proliferation and survival of CRC cells. Towards HCT116 CRC cells. Importantly, loss of SNHG5 reduces the identifying a function for SNHG5 we profiled its expression in a protein levels of the interactors via destabilization of their panel of CRC cell lines (Supplementary Fig. 3a) and designed two mRNAs. We further characterize the interaction of SNHG5 independent short interfering RNAs (siRNAs) targeting SNHG5 with the target SPATS2, and demonstrate that loss with knockdown efficiencies 475% (Fig. 1f). The cytosolic frac- of SPATS2 phenocopies the effect of SNHG5 depletion. STAU1 tion of the transcript was the most strongly reduced when is a part of a highly conserved family of double-stranded analysed 36 h after siRNA transfection (Supplementary Fig. 3b). RNA-binding proteins implicated in mRNA transport, stability Moreover, the expression levels of the U50A snoRNA remained and translation14–17. We show here that SNHG5 binding to target unchanged, suggesting that the siRNAs target the spliced tran- mRNAs protects these from STAU1-mediated degradation. script and that phenotypic effects will be independent of the Importantly, loss of STAU1 also rescues the apoptotic effect snoRNAs, which in these cell lines exerts a nuclear restricted of SNHG5 depletion. expression (Supplementary Fig. 3c,d). To investigate biological Hence, we here provide new insights into the significance of processes affected by SNHG5 depletion, RNA-seq was performed lncRNAs in CRC in general and to the specific role of SNHG5 in in HCT116 cells 36 h after transfection with two indepen- promoting CRC cell survival. dent siRNAs against SNHG5 or a negative control siRNA

2 NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875 ARTICLE

ab 1,500 1,000 SNHG5 SNHG5 **** 800 *** 1,000 600 **** FPKM

FPKM 400 500 200

0 0 N T N A I II III IV (n=294) (n=313)

c 2 kb chr6: 86, 387, 000 86, 388, 500

SNHG5-01 < < SNHG5-02 < < < U50A U50B

deSNHG5 U50A ACTB 3 3 3 SNHG5 DAPI Merged 2 2 2

1 1 1 (HCT-116) HCT-116 HCT-116

Relative expression 0 0 0 C N CN C N 3 3 3

2 2 2

1 1 1 TIG-3 fibroblasts (TIG-3 fibroblasts) Relative expression 0 0 0 CN CN CN

fg1.5 h

siSNHG5 1 siSNHG5 2 1.0 Protein modification process GAPDH Developmental process

expression Metabolic process 0.5 920 150 490 FC> ±3 Cell communication **

relative to Cellular process SNHG5 ** 0.0 0 2 4 6 – Log2 P value FDR < 0.05

siControl siSNHG5siSNGH5 1 2 Figure 1 | SNHG5 is deregulated in CRC. (a) The expression levels of SNHG5 were proﬁled in clinical specimens by RNA-seq, where the RNA was extracted from tumour (C) or adjacent normal tissues (N) (****P valueo0.0001, Mann–Whitney t-test). (b) Box plot representing stratiﬁed SNHG5 expression in CRC clinical specimens (N ¼ normal adjacent tissue, A ¼ adenomas, I–IV ¼ CRC progression stages). The bottom line of each box represents the 1st quartile of sample population, the top line of the box represents the third quartile, while the middle quartile represents the population median. (***P valueo0.001, ****P valueo0.0001, Mann–Whitney t-test). (c) Schematic representation of the SNHG5 locus. (d) RNA-FISH of SNHG5 in red (Cy3) with a nuclear DAPI stain (blue) in HCT116 and TIG-3 cells. Scale bar, 50 mm. (e) RNA from the nuclear (N) and cytosolic (C) fractions were isolated from HCT116 and TIG-3 cells and SNHG5 levels measured using qRT–PCR. U50A and ACTB were used as nuclear and cytosolic controls, respectively. The expression values for each transcript are relative the total RNA input and normalized on the nuclear fraction expression levels. Error bars represent s.e. from one representative experiment. (f) qRT–PCR. HCT116 cells were transfected with 50 nM siSNHG5-1-2 or control siRNA and the RNA collected after 36 h. SNHG5 expression data are showed normalized to the GAPDH housekeeping gene and normalized to the non-targeting control siRNA. (g) Venn diagram representing the number of genes altered in HCT116 cells following siSNHG5-1-2 transfection after 36 h (log2 fold change4 ±3, FDRo0.05). (h) Gene ontology annotation. Statistical overrepresentation test (Bonferroni-corrected) was applied to identify gene categories affected by SNHG5 depletion.

(Fig. 1g). We found 150 genes to be significantly deregulated by SNHG5 depletion impacts key survival pathways in CRC cells, both SNHG5 siRNAs (fold change 43, FDRo0.05). Gene such as the STAT3 pathway (Supplementary Fig. 3e). This was ontology analysis revealed a significant enrichment of genes further confirmed using western blots for STAT3, a known involved in differentiation processes, such as mesoderm devel- STAT3 target gene BCL-XL, and BCL-2 (Supplementary Fig. 3f). opment, cell communication and metabolic processes (Fig. 1h). Importantly, knockdown of SNHG5 (Supplementary Fig. 3g) Furthermore, gene set enrichment analysis20 suggested that resulted in a strong anti-proliferative effect in CRC cells (Fig. 2a).

NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875

This result was confirmed by cell cycle analysis using EdU As siRNA depletion studies may suffer from off-target effects, incorporation, which revealed a significant decrease in the we subsequently investigated the effect of SNHG5 overexpression. S-phase cell population in SNHG5-depleted cells (Fig. 2b). To Towards this, both a cDNA version (430 bp) and the entire investigate an impact of SNHG5 on survival, CRC cell lines were genomic locus of SNHG5 (1800 bp) were cloned (Fig. 2d) stained for cleaved Caspase-3 protein and analysed by flow and transiently expressed in HT-29 cells, which display a cytometry, which revealed a strong induction of apoptosis in both lower endogenous SNHG5 level as compared with HCT116, HCT116, CACO-2 and DLD-1 cells (Fig. 2c, top). This was CACO-2 and DLD-1 cells (Supplementary Fig. 3a). Forty-eight confirmed by western blots for cleaved PARP-1 (Fig. 2c, bottom). hour after transfection a 60-fold average increase in the SNHG5

a b 5 HCT-116 5 CACO-2 50 HCT-116 4 ** 4 CACO-2 3 ** 3 * 40 DLD-1 2 2 1 1 30 * *

Relative growth 0 0 Time (h) 24 48 72 24 48 72 20 * * 5 DLD-1 10 * % EdU positive cells ** 4 * siControl 3 siSNHG5 1 0 2 siSNHG5 2 siControl + + + 1 siSNHG5 1 + + + Relative growth 0 siSNHG5 2 + + + Time (h) 24 48 72 c d 60 HCT-116 1.8 kb ** CACO-2 pCDNA 3.1 SNHG5 gLOCUS 40 DLD-1 430 bp ** pCDNA 3.1 SNHG5 cDNA 20 ** * positive cells * e

% Cleaved caspase-3 0 HT-29 HT-29 (oxaliplatin 25 μM) siControl + + + 80 60 siSNHG5 1 + + + siSNHG5 2 + + + 60 40 HCT-116 CACO-2 DLD-1 GAPDH

expression 40 - 98 c-PARP-1 20 ** 20 *** GAPDH - 38 positive cells

SNHG5 relative to 0 0

% Cleaved caspase-3 siControl + + + pEmpty + pEmpty + siSNHG5 1 + + + pSNHG5 cDNA + pSNHG5 cDNA + siSNHG5 2 + + + pSNHG5 Locus + pSNHG5 Locus +

fg 60 WT SNHG5 Δ 1–40 (exon 1) 40 Δ 41–93 (exon 2) * ** Δ 94–172 (exon 3) ** Δ 173–253 (exon 4) 20 **

Δ 254–322 (exon 5) positive cells Δ 323–430 (exon 6) % Cleaved caspase-3 0

Empty cDNA

1–40 (exon 1) Δ 41–93 (exon 2) pSNHG5 Δ 94–172 (exon 3) Δ 173–253254–322 (exon323–430 (exon4) (exon 5) 6) Δ Δ Δ h i 11 2.0×10 shSNHG5-4 – DOX 2,000 – DOX 2,000 ) P=0.369 + DOX 11 3 1.5×10 shGFP 1,500 1,500 P=0.0094 1.0×1011 P=0.795 1,000 1,000 5.0×1010

Photons/s 0.0 500 500 Volume (mm Volume –5.0×1010 0 0 Time (days) 1 7 13 19 25 29 35

2.0×1011 shSNHG5-4 + DOX j 2,000 P=0.5973 2,000 1.5×1011 shGFP – DOX + DOX 11 1,500 1,500 P=0.0211 1.0×10 P=0.017 (mg) 5.0×1010 1,000 1,000

Photons/s 0.0 500 500 Weight –5.0×1010 0 0 Time (days) 1 7 13 19 25 29 35 shGFP shSNHG5-4

4 NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875 ARTICLE levels were detected by qRT–PCR in comparison to the empty specific oligonucleotide pool was used as negative control vector control (Fig. 2e, left). The cells were subsequently treated (Fig. 3a; Supplementary Fig. 5a). The pull-down efficiency was for 24 h with oxaliplatin, a drug commonly employed in CRC estimated by qRT–PCR to B50% of the endogenous SNHG5 treatment. Interestingly, using flow cytometry, we observed a transcript pool in HCT116 cells (Fig. 3b). The interacting significant decrease in the cleaved Caspase-3 positive population RNA transcripts were identified using single-end 50 nucleotide of HT-29 cells overexpressing SNHG5 (Fig. 2e right). This Illumina sequencing and the sequence reads were mapped protective effect was evident for both the cDNA construct and the onto the Hg19 genome scaffold using the STAR aligner tool entire SNHG5 locus, suggesting no significance of the snoRNAs in (see ‘Methods’ section for details). The relative enrichment of this assay. No proliferative effect was observed in HT-29 cells peaks present in both the Even and the Odd pools were calculated overexpressing the SNHG5 lncRNA in the absence of drug with respect to the luciferase pool, which was used as back- treatment (Supplementary Fig. 4a,b). Next, we created a panel of ground reference. We identified 121 significant peaks overla- SNHG5 mutants lacking individual exons and tested the ability of pping between the Even and Odd pools (P valueo0.05) these mutants to protect HT-29 cells from oxaliplatin-induced (Fig. 3c; Supplementary Fig. 5b). About half of the peaks apoptosis (Fig. 2f). Interestingly, whereas mutants for exons 2, 3, mapped to coding regions (51.4%) while the majority of the 5 and 6 all protected the cells to the same extend as full-length remaining peaks resided in 30 UTR regions (40%) (Supplementary SNHG5, mutants lacking either exon 1 or exon 4 failed to protect Fig. 5c). against oxaliplatin-induced apoptosis (Fig. 2g). We ranked a short list of SNHG5-interacting mRNAs based We next employed a luminescent xenograft mouse model on the number of peaks (overlapping and not-overlapping) in to query whether SNHG5 expression is required for tumour each transcript and their local enrichment relative to growth in vivo (Supplementary Fig. 4c). Luminescent HCT116 the luciferase background pool (Supplementary Fig. 5d). The CRC cells holding doxycycline-inducible shRNAs targeting mRNA transcripts representing the peaks were subsequently SNHG5 or GFP as control were subcutaneously injected into validated in HCT116 by qRT–PCR using peak-specific primers two groups of immune-compromised mice. To allow time for (Fig. 3d; Supplementary Fig. 5d,e). We identified the SPATS2, tumour establishment, doxycycline was added to the drinking PITRM1 and GLE1 mRNAs as the most enriched SNHG5- water at day 13 after injection. SNHG5 levels were significantly interacting transcripts. We further validated the RIA-seq data reduced in the tumours following doxycycline treatment comp- in CACO-2 and DLD-1 cells, where SNHG5 depletion was also ared with the non-targeting control (Supplementary Fig. 4d). found to affect cell survival (Fig. 2a). In these cells only the Importantly, we observed a significant negative effect on tumour interaction with the SPATS2 mRNA could be consistently growth following SNHG5 depletion (Fig. 2h; Supplementary confirmed, probably due to lower expression levels of the other Fig. 4e). SNHG5 depletion resulted in a significant reduction of target transcripts in these cell lines (Supplementary Fig. 6a,b). tumour volume and weight as measured at the end of the Importantly, SNHG5 depletion results in a clear decrease of experiments at day 35 (Fig. 2i,j). Altogether, this data suggest that the target protein levels (Fig. 3e). To evaluate the effect of SNHG5 SNHG5 expression functions to regulate proliferation and on the stability of the target transcripts, HCT116 cells survival of CRC cells both in vitro and in vivo. were treated for 5 h with the RNA PolII inhibitor Triptolide following SNHG5 knockdown (Fig. 3f). The mRNA stabilities of all three top interactors were reduced on SNHG5 knockdown as RIA-seq identifies primary downstream effectors of SNHG5. quantified using qRT–PCR. Altogether, this data suggest that Until now, only few mechanisms for cytoplasmic lncRNAs have SNHG5 forms RNA–RNA interactions with target mRNAs in the been revealed. To unveil the mechanism by which SNHG5 affects cytoplasm and regulates their stability. proliferation and apoptosis in the cytoplasm, we used an unbiased method, RIA-seq (RNA interactome analysis and sequencing), to map transcriptome-wide RNA–RNA interactions21. Two sets SPATS2 is a key downstream target of SNHG5. To further (Even and Odd) of eleven 30-biotinylated DNA oligonucleotides explore the underlying mechanism, we focused on the SNHG5– were designed to specifically hybridize to SNHG5. A luciferase- SPATS2 interaction and confirmed that SNHG5 depletion in

Figure 2 | Manipulation of SNHG5 expression regulates proliferation and survival both in vitro and in vivo. (a) Proliferation assay. CRC cell lines were transfected with 25 nM siSNHG5-1-2 or siControl. Cell numbers were assessed at the designated time points post transfection. Data normalized to the first time point are shown for the mean of 3 independent experiments. Error bars indicate±s.d. (*P valueo0.05, **P valueo0.01, Student’s t-test paired). (b) HCT116, CACO-2 and DLD-1 cells were transfected as in a, pulsed with EdU after 36 h and analysed by flow cytometry to measure the number of EdU- labelled cells representing the S-phase populations. Data represents the mean of three independent experiments. Error bars indicate±s.e. (*P valueo0.05, **P valueo0.01). (c) Top panel: CRC cell lines were transfected as in (a). Thirty-six hours after transfection, cells were labelled with cleaved Caspase-3 antibody and analysed by flow cytometry. Data are shown for the mean of three independent experiments. Error bars indicate±s.e. (*P valueo0.05, **P valueo0.01). Bottom panel: western blot analysis of endogenous cleaved PARP-1 levels in HCT116, CACO-2 and DLD-1 cells transfected with siSNHG5-1-2 or siControl. GAPDH is included as loading control. A representative experiment is shown (n ¼ 3). (d) Schematic representation of constructs expressing the entire SNHG5 genomic (pSNHG5 locus) or only the cDNA (pSNHG5 cDNA). (e) Left, qRT–PCR. SNHG5 expression levels were measured 48 h after transfection into HT-29 cells. Data was normalized to the RPLP0 housekeeping gene and plotted relative to the empty vector control. Error bars indicate the mean±s.e. of three independent experiments. Right, HT-29 cells were transfected with plasmids as described above. Forty-eight hours after transfection the cells were treated with 25 mgml 1 of oxaliplatin. Twenty-four hours later the cells were stained with cleaved Caspase-3 antibody and analysed by flow cytometry. Error bars indicate the mean±s.e. of three independent experiments. (*P valueo0.05, **P valueo0.01, ***P valueo0.01). (f) Schematic representation of the SNHG5 cDNA expressing mutant plasmids. (g) Overall, 2 105 HT-29 cells were transfected with mutant plasmids and treated with oxaliplatin as described above. Twenty-four hours later the cells were stained with cleaved Caspase-3 antibody and analysed by flow cytometry. Error bars indicate the mean±s.e. of three independent experiments. (*P valueo0.05, **P valueo0.01). (h) Xenograft growth of HCT116 pFULT-shGFP and HCT116 pFULT-shNHG5-4 cells. Doxycycline was added to the drinking water after 13 days. Mean±s.d. of tumor luminescence (photons/seconds/cm2) at the indicated time points (for statistic analysis details see ‘Methods’ section). (i,j) Final volume and weight of tumours from the experiment at day 35, respectively.

NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875

a b 0.8 SNHG5

0.6 GAPDH

0.4 SNHG5 < < < 0.2 Retrieved RNA/input Retrieved 0.0 Even + + Odd + + Luciferase + + cd SPATS2 0.8 PITRM1 GLE1 Even Odd IRX4 0.6 ZNF335 CPNE1 0.4 1,334 121 905 SOCS6 KIA00141 0.2

Retreived RNA/input Retreived CCD53 P value <0.05 0.0 Even + + + + + + + + + Odd + + + + + + + + + Luciferase + + + + + + + + +

e f 1.5 OD 1 0.68 0.79 SPATS2 PITRM1 -98 GLE1 100 OD 1 0.66 0.67 PITRM1 1.0 -98 80

GAPDH GLE1

expression * 60 * * 0.5 OD 1 0.51 0.48 * * SPATS2 - 62 40 *

RNA (%) SNHG5 relative to 20 0.0 GAPDH - 38 Fraction of remaning 0 siControl + + + siControl siSNHG5 1 + + + siSNHG5siSNGH5 1 2 siControl siGLE1 siSNHG5 2 + + + siPITRM1 siSNHG5siSNGH5 1 2 siSPATS2

Figure 3 | SNHG5 targets mRNAs in the cytoplasm. (a) Schematic representation of biotinylated antisense DNA probes complementary to the SNHG5 transcript Even (green) and Odd (red) used for the RNA pool-down. (b) qRT–PCR was performed to assess the % of SNHG5 RNA retrieved from HCT116 cell lysates following pull-down with streptavidin beads. Data was normalized to input (1:100) and GAPDH mRNA included as negative control. Error bars indicate±s.d. for the mean of three independent experiments. (c) Venn diagram representing the number of SNHG5 pull-down peaks commonly enriched with both the Even and Odd probe sets. (d) Validation of the RIA-seq results by qRT–PCR. Data for each target transcript is normalized on input (1:100). Error bars indicate±s.d. for the mean of three independent experiments. (e) Left, qRT–PCR. SNHG5 expression levels were measured 48 h after transfection of HCT116 cells. Data was normalized to the GAPDH housekeeping gene and plotted relative to the siRNA control. Error bars indicate the mean±s.e. of three independent experiments. Right, western blot. HCT116 cells 48 h after transfection with the indicated siRNAs. The optical density (OD) of protein bands is indicated relative to corresponding loading control GAPDH and normalized relative the non-targeting siControl. (f) Transcript stability of SNHG5 targets were measured by qRT–PCR in HCT116 cells 36 h after transfection with the indicated siRNAs. The cells were treated with Triptolide at a ﬁnal concentration of 10 mM for the last 5 h. The mRNA levels are plotted relative to the corresponding expression levels in the untreated cells. Error bars indicate the mean±s.e. of three independent experiments (Student’s t-test paired).

HCT116, CACO-2 and DLD-1 cells, where both transcripts are (PETcofold) point towards these regions as the interaction site. expressed at similar levels (Supplementary Fig. 6c), results in a The interaction is only conserved among primates (phylogenetic decrease of the SPATS2 protein levels (Fig. 4a). The SPATS2 gene tree predicted in PETcofold) and the phyloP base-wise product is a predicted cytoplasmic RNA-binding protein, but conservation (100 vertebrates UCSC genome alignment) shows besides its potential involvement in the maturation of gonad cells low evolutionary pressure on SNHG5 suggesting an evolutionary SPATS2 functions are unknown22. Importantly, SNHG5 over- new binding site (Fig. 4c). As expected, overexpression of SNHG5 expression in HT-29 was found to increase SPATS2 protein levels promotes the stabilization of the SPATS2 mRNA (Supplementary (Fig. 4b). An RNA–RNA interaction site was computatio- Fig. 6d). To identify the region in SNHG5 responsible for nally predicted to span the ﬁrst exon of SNHG5 and the promoting SPATS2 mRNA stability, we overexpressed the panel 30 UTR of SPATS2. Both the thermodynamics prediction of SNHG5 mutants and measured the stability of the SPATS2 (Energy ¼15.50 kcal mol 1; IntaRNA) and the conservation mRNA following Triptolite treatment. Interestingly, the SNHG5

6 NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875 ARTICLE

a HCT-116 CACO-2 DLD-1 b HT-29 40 OD 1 1.45 OD 1 1 1 0.45 0.52 0.65 0.69 0.26 0.15 30 - 62 - 62 SPATS2 GAPDH SPATS2 20

expression

GAPDH 10 GAPDH - 38 - 38

SNHG5 relative to 0 siControl + + + Empty + Empty + siSNHG5 1 + + + siSNHG5 2 + + + pSNHG5 cDNA + pSNHG5 cDNA +

c - 15.5 Kcal mol–1 SNHG5 5′ .... 3′

3′...... 5′ SPATS2 3′UTR PETcofold 4 IntaRNA PhyloP conservation hg19 –4 panTro2 ponAbe2 rheMac2 papHam1 gorGor1

d e SPATS2 3′ UTR CMV LUCIFERASE 2456-2660

6 SPATS2 pReport-SPATS2bs 100 * * * * pReport-SPATS2bs competed 5 80 pReport control 4 pReport control competed 60 n.s. n.s. 2 40

Firefly/ renilla ratio 20 1

Fraction of remaining RNA (%) 0 0 Empty + + + +

Empty pSNHG5 cDNA + + + + Δ1–40 (exon 1) + + + + 1–40 (exon 1) pSNHG5Δ cDNA Δ 41–9394–172 (exon 2)(exon 3) Δ 173–253254–322 (exon323–430 4) (exon 5) (exon 6) Δ Δ Δ

Figure 4 | SNHG5 promotes SPATS2 mRNA stability. (a) Western blot. HCT116, CACO-2 and DLD-1 cells were transfected with the indicated siRNAs and proteins isolated after 48 h. OD of protein bands is indicated relative to corresponding loading control GAPDH and normalized relative the non-targeting siControl. (b) Left, 2 105 HT-29 cells were transiently transfected with the indicated plasmids. SNHG5 expression levels were measured 48 h after transfection into HT-29 cells using qRT–PCR. Right, western blot. SPATS2 expression levels were evaluated 48 h after transfection. OD of protein bands is indicated relative to corresponding loading control GAPDH and normalized relative the empty vector. (c) For graphical output of the interaction between SNHG5 exon 1 and the SPATS2 30 UTR were combined PETcofold analysis (red) relative to the evolutionary conservation of the interaction site and IntRNA (green) to map the most thermodynamically stable interaction between the lncRNA and the mRNA. Local PhyloP conservation index is included. (d) qRT–PCR. 2 105 HT-29 cells were transfected with the indicated plasmids. Forty-eight hours after transfection the cells were treated with Triptolide at a final concentration of 10 mM for 8 h and the RNA subsequently extracted. The percentage retrieved SPATS2 mRNA was obtained normalizing to the corresponding expression levels in the untreated cells. (e) Top, schematic representation of the 30-UTR region of the SPATS2 30 UTR fragment cloned in the pMIR-REPORT plasmid. Bottom, luciferase reporter assay 48 h after transfection of 2.4 104 HEK293 cells per well (four well each samples) with the indicated plasmids and a renilla luciferase transfection control plasmid. To compete the SNHG5 binding with the SPATS2 30-UTR, pcDNA 3.1 expressing the complete SPATS2 cDNA was co-transfected with the indicated plasmids. Error bars indicate±s.e. of three independent experiments (*P valueo0.05). mutant lacking exon 1 failed to promote SPATS2 mRNA stability to compete the SNHG5 interaction with a second luciferase (Fig. 4d), thus sustaining the notion that SNHG5 interacts with construct containing the SNHG5 binding site from the GLE1 SPATS2 via elements in SNHG5 exon 1. 30-UTR (nts 2,496–2,706) (Supplementary Fig. 6e,f). Altogether, We then assessed the binding of SNHG5 to the 30-UTR of these data demonstrate that SNHG5 can directly bind and SPATS2 (nts 2,456–2,660) using a luciferase reporter system regulate SPATS2. (Fig. 4e, top). The ectopic expression of SNHG5 in HEK293 cells increased the luciferase activity of the SPATS2 30 UTR reporter construct, as compared with the control, whereas the SNHG5 The SNHG5–SPATS2 axis promotes CRC cell survival. SPATS2 mutant lacking exon 1 was unable to do so (Fig. 4e, bottom). In has not previously been implicated in cancer-related processes, addition, we were able to compete the SNHG5 binding to the however, RNA-seq data analysis of CRC-paired samples showed SPATS2 30-UTR by co-transfecting the cells with a plasmid a significant up-regulation of the SPATS2 transcript in tumours vector expressing the full-length SPATS2 mRNA. In addition, versus correspondent normal tissues (Fig. 5a). Similarly, we co-transfecting the full-length SPATS2 mRNA was also sufficient find the two other top interactors GLE1 and PITRM1 to be

NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875

a b 6 **** 60 HCT-116 CACO-2 * DLD-1

4 40

Log2 (FPKM+1) 2 positive cells 20 *

%Cleaved caspase-3

0 0 NTsiControl + + + (n = 268) siSPATS2 + + +

cd 30 P < 0.0001 00.4

- 62 SPATS2 20 00.3

HPRT1 00.2

* expression 10

positive cells GAPDH - 38 00.1

% Cleaved caspase-3 relative to relative

0 SPATS2 Lenti-control + Lenti-control + 0.0 Lenti-SPATS2 + Lenti-SPATS2 + shGFP (DOX) + shSNHG5-4 (DOX) +

efg 60 ** 60 OD 1 ** ** * 0.46 0.39 3.59 4.23 4.21 40 * 40 ** * - 62 * SPATS2 20 20

positive cells GAPDH - 38

%Cleaved caspase-3 0 % EdU positive cells 0 siRNA control + + siRNA control + + siRNA control + + siRNA SNHG5-1 + + siRNA SNHG5-1 + + siRNA SNHG5-1 + + siRNA SNHG5-2 + + siRNA SNHG5-2 + + siRNA SNHG5-2 + + Lenti-SPATS2 +++ Lenti-SPATS2 + + + Lenti-SPATS2 +++

Figure 5 | The SNHG5–SPATS2 axis regulates viability in CRC cells. (a) Scatter plot. The expression levels of SPATS2 were profiled in paired CRC clinical specimens by RNA-seq, where the RNA was extracted from either tumor (C) or normal tissues (N) (****P valueo0.0001, Mann–Whitney t-test). (b) HCT116, CACO-2 and DLD-1 cell lines were transfected with indicated siRNAs. Thirty-six hours after transfection, cells were labelled with cleaved Caspase-3 antibody and analysed by flow cytometry. Data are shown for the mean of three independent experiments. Error bars indicate±s.e. (*P valueo0.05). (c) Left, western blot. SPATS2 expression levels were measured 48 h after stable transduction of HT-29 cells. GAPDH is included as loading control. Right, HT-29 cells the cells were treated with 25 mgml 1 of oxaliplatin. Twenty-four hours later, the cells were stained with cleaved Caspase-3 antibody and analysed by flow cytometry. Error bars indicate the mean±s.e. of three independent experiments. (*P valueo0.05). (d) qRT–PCR. SPATS2 mRNA levels in shGFP and shSNHG5-4 xenograft tumours at day 35. Data is plotted relative to the HPRT1 housekeeping gene (****P valueo0.0001). (e) Western blot. HCT116 stable transduced with the indicated lentiviral vectors were transfected with indicated siRNAs. SPATS2 expression levels were measured 36 h after transfection. OD of protein bands is indicated relative to corresponding loading control GAPDH and normalized relative the non-targeting siControl. (f) HCT116 cells were transfected as in e, stained with cleaved Caspase-3 antibody or (g) pulsed with EdU after 36 h to measure the number of EdU-labelled cells representing the S-phase populations and analysed by flow cytometry. Data represents the mean of three independent experiments. Error bars indicate±s.e. (*P valueo0.05, **P valueo0.01). up-regulated in CRC (Supplementary Fig. 7a). If the stabilizing induction on oxaliplatin treatment, thus recapitulating the effect of SNHG5 on the SPATS2 mRNA is important for the SNHG5 overexpression phenotype (Fig. 5c). survival-promoting functions of SNHG5, we would expect loss of To further explore the interaction between SNHG5 and SPATS2 to impact on CRC cell survival. Indeed, knockdown of SPATS2, we investigated the SNHG5 and SPATS2 expression SPATS2 in CRC cell lines resulted in a significant reduction of cell patterns in primary CRCs and mouse xenograft tumours viability (Fig. 5b; Supplementary Fig. 7b). Furthermore, pheno- (Supplementary Fig. 7d). The correlation found in CRC samples copying the effects observed on SNHG5 depletion, also the was further confirmed by analyzing SPATS2 levels in inhibition of SPATS2 expression down-regulates the STAT3 the xenograft tumours, where knockdown of SNHG5 resulted pathway, hence confirming the impact of SPATS2 expression on in a significant decrease in SPATS2 mRNA levels (Fig. 5d). viability in CRC cell lines (Supplementary Fig. 7c). Moreover, Finally, we investigated whether restoring SPATS2 expression ectopic SPATS2 expression in HT-29 cells prevented apoptosis in SNHG5-depleted CRC cells was sufficient to rescue the

8 NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875 ARTICLE apoptotic phenotype. Indeed, in HCT116 cells ectopic expre- U50 and U500 in introns 4 and 5, respectively. Whereas snoRNAs ssion of SPATS2-mediated a partial rescue from the anti- of the C/D box family are well-known to guide fibrillarin- proliferative and pro-apoptotic effect following the SNHG5 mediated 20O methylation of the ribosomal RNA, reduced depletion (Fig. 5e–g). Altogether, this data suggest that the expression and both somatic and germ line mutations in U50 SNHG5–SPATS2 interaction is important for promoting CRC cell have been reported from breast and prostate cancer30,31. survival. Furthermore, a non-canonical function of the U50 snoRNA in modulating RAS activity was recently described32. Although currently unclear, this is likely independent from the SNHG5 SNHG5 impairs the association of SPATS2 with STAU1. mechanism characterized in this paper. We hypothesized that the binding of SNHG5 promoted SPATS2 We find that SNHG5 expression levels are significantly elevated mRNA stabilization via protecting the transcript from post- both in the transition between normal tissue and adenomas transcriptional regulators. The RNA-binding protein STAU1 has and between adenomas and early stage (I) carcinoma. Impor- previously been described to destabilize mRNAs in the cyto- 23 tantly, suppression of SNHG5 expression not only exerts a strong plasm . Interestingly, knocking down STAU1 resulted in an anti-proliferative effect and induces apoptosis, impacting key up-regulation in both transcript and protein levels of SPATS2 in survival pathways such as the STAT3 pathway in vitro, but also HCT116 cells (Fig. 6a). To evaluate the effect of STAU1 depletion reduces tumour progression in vivo. Moreover, SNHG5 over- on SPATS2 transcript stability and to compare with the effect expression prevents apoptosis induction in CRC cell lines treated resulting from SNHG5 depletion, HCT116 cells were treated for with chemotherapeutic agents. We further characterize SNHG5 as 5 h with Triptolide following SNHG5 and STAU1 knockdown a stable, cytoplasmic lncRNA not associated with polysomes. This (Fig. 6b). As expected, the SPATS2 mRNA stability was reduced prompted us to explore mechanisms for SNHG5 in the cytoplasm on SNHG5 depletion, whereas the opposite effect was measured by isolating mRNAs interacting with the lncRNA using RIA-seq. after STAU1 knockdown. To validate SPATS2 as a direct target of This is in line with an increasing number of studies describing STAU1, we performed RNA immunoprecipitation using a post-transcriptional mechanisms of gene expression regulation STAU1 antibody after ultraviolet cross-link. Importantly, the mediated by cytosolic lncRNAs22. We identified 121 significant SPATS2 mRNA was enriched in the STAU1 pull-down compared sites of intermolecular interaction with SNHG5, located mainly in with the IgG control. Moreover, depleting SNHG5 in HCT116 cis-regulatory sequences (30 UTRs), and further pinpointed cells increased the association of SPATS2 with STAU1. Impo- SPATS2 as an important direct target of SNHG5 in multiple rtantly, this was not observed for another known STAU1 target CRC cell lines. SPATS2 is significantly up-regulated in CRC gene, ARF1 (Fig. 6c)13. This data suggests that SNHG5 impairs patients and knockdown of SNHG5 destabilizes the SPATS2 the association SPATS2 with STAU1. Having confirmed SPATS2 mRNA and significantly reduces SPATS2 protein expression. as a key target of SNHG5, and STAU1 as a regulator of SPATS2, Vice versa, overexpression of SNHG5 leads to increased SPATS2 we speculated if STAU1 could be implicated in CRC cell survival. mRNA stability and protein expression. In vitro luciferase Towards this, we combined knockdown of STAU1 with reporter assays demonstrated that this effect is mediated via oxaliplatin treatment and measured the level of apoptosis. direct interaction between a sequence motif in the SNHG5 exon As evident from Supplementary Fig. 7e, loss of STAU1 reduced 1 and the 30 UTR of SPATS2, although we cannot rule out other the level of apoptosis implicating STAU1 in CRC cell survival. interactions modes with other targets or in other cellular contexts. We queried the pathway hypothesizing that if SNHG5 and Importantly, manipulating the expression levels of SPATS2 STAU1 have opposite functions on SPATS2 mRNA stability, in CRC cells phenocopies the cellular viability response loss of STAU1 might rescue the apoptosis seen following SNHG5 following SNHG5 depletion or overexpression. Finally, restoring loss. Towards this, we combined the knockdown of SNHG5 and the expression of SPATS2 in SNHG5-depleted cells resulted in STAU1 and measured the resultant level of apoptosis in HCT116 a partial rescue of the apoptotic phenotype induced by the cells. As evident from Fig. 6d,e, loss of STAU1 expression strongly inhibition of lncRNA. Hence, SPATS2 is a key downstream target reduced the apoptosis induction following SNHG5 depletion. of SNHG5. Interestingly, we find that the expression levels In conclusion, our analyses demonstrate that SNHG5 promotes of SNHG5 and SPATS2 display a significant negative correlation CRC cells survival by alleviating STAU1-induced degradation in adenomas and xenograft tumours following SNHG5 of SPATS2. knockdown, suggesting a contribution of the SNHG5–SPATS2 axis to tumorigenic events in CRC. Discussion Several RNA-binding proteins have been described to regulate In this study, we used a customized array platform to profile the mRNA stability and decay33. Here, we demonstrate in CRC cells expression of lncRNAs in a set of CRC samples and to identify that the SNHG5–SPATS2 interaction impairs SPATS2-association lncRNA expression profiles correlating with disease progression. with STAU1, a major regulator of the cytoplasmic RNA decay. We identified SNHG5 as significantly up-regulated in CRC with STAU1 is part of a highly conserved family of double-stranded respect to normal tissues and validated this expression pattern RNA-binding proteins implicated in mRNA transport, stability in an independent cohort of 313 CRCs. SNHG5 belongs to a and translation; these processes depend on the ability of STAU1 large family of non-coding genes hosting small RNAs, such as to bind intra- or intermolecular RNA duplexes, most prominently snoRNAs and microRNAs, most often residing in the introns of found in 30 UTRs of target mRNAs, and to recruit components of the host genes. Following excision of the small RNAs, these host the cytosolic decay machinery, such as UPF1 (ref. 13). STAU1 is gene transcripts in general are unstable24, and many have been ubiquitously expressed, and whereas an important role for viewed as biologically inert carriers, where the small RNAs are STAU1 in the development of the central nervous system has thought to be the biologically relevant players25. Noticeable been described, the involvement of STAU1 in tumorigenic exceptions include host genes such as GAS5, TINCR, ncRAN processes remains largely unexplored. STAU1 has previously and MIR31HG21,26–28, all of which have biological effects been linked to lncRNA functions as a post-transcriptional independent of their hosted small RNAs. SNHG5 has not previ- regulator of the stability of specific mRNAs involved in ously been characterized in CRC but was reported as a epidermal terminal differentiation21,34 and cell cycle entry in translocation partner to BCL6 in a patient suffering from gastric cancer cell lines35. We analysed a recently published diffuse large B-cell lymphoma29. SNHG5 encodes the snoRNAs hiCLIP dataset for STAU1 (ref. 17) and did not find such sites

NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875

a b HCT-116 2.0 100 OD 1 0.09 * * STAU1 80 - 49 1.5

GAPDH - 38 GAPDH 60 * * expression 1.0

remaining mRNA 40 OD 1 1.80 0.5 - 62 20

SPATS2 SPATS2 relative to 0.0 0.0

GAPDH - 38 % SPATS2

siControlsiSTAU1 siControl siSTAU1 siSNHG5-1siSNHG5-2 siControlsiSTAU1

c 40 ARF1 siControl ARF1 siSNHG5 30 SPATS2 siControl * SPATS2 siSNHG5 20 H3 siControl H3 siSNHG5 10 Enrichment relative to INPUT to Enrichment relative 0.0

IgG

Anti-STAU1

d e OD 1 0.01 0.98 0.02 0.99 0.01 50 STAU1 - 49 40 * * GAPDH - 38 30

OD 1 20 0.51 0.57 3.59 1.85 2.05

positive cells ** SPATS2 - 62

% Cleaved caspase-3 10

GAPDH - 38 0 siControl + siControl + siSNHG5 1 + + siSNHG5 1 + + + siSNHG5 2 + + siSNHG5 2 + siSTAU1 + + + siSTAU1 + + + Figure 6 | SNHG5 impairs the association between STAU1 protein and the SPATS2 mRNA. (a) Left, western blot for STAU1 and SPATS2 in HCT116 cells transfected with indicated siRNAs for 72 h. OD of protein bands is indicated relative to corresponding loading control GAPDH and normalized relative the non-targeting siControl. Right, qRT–PCR. HCT116 cells were transfected as described above. SPATS2 expression data are shown normalized to the GAPDH housekeeping gene and normalized to the non-targeting control siRNA. (b) qRT–PCR. HCT116 cells were transfected with the indicated siRNAs. Forty-eight hours after transfection the cells were treated with Triptolide at a final concentration of 10 mM for 5 h and the RNA subsequently extracted. The percentage retrieved SPATS2 mRNA was obtained by normalizing to the corresponding expression levels in the untreated cells. (c) RNA immunoprecipitation. HCT116 were transfected with the indicated siRNAs for 48 h. The cytosolic fraction was isolated from the ultraviolet-cross-linked cells and the IP performed with a STAU1-specific antibody. Rabbit IgG were included as negative control for the immunoprecipitation. The RNA was extracted, retro-transcribed and the SPATS2 mRNA levels evaluated. ARF1 was included as positive control and H3 mRNA as negative control. Error bars indicate±s.d. relative for the mean of three independent experiments (*P valueo0.05, stats method). (d) Left panel, western blot. HCT116 cells were transfected with indicated siRNAs for 72 h and analysed for STAU1 and SPATS2 using western blotting. OD of protein bands is indicated relative to corresponding loading control GAPDH and normalized relative the non-targeting siControl. Right panel, Flow Cytometry. (e) HCT116 cells were transfected as in d, the cells stained with cleaved Caspase-3 antibody, and analysed by flow cytometry. Error bars indicating±s.d. relative for the mean of three independent experiments (*P valueo0.05, **P valueo0.01). in SPATS2. This, however, is unsurprising as that study was STAU1-mediated mRNA decay and that SNHG5 can limit performed in HEK293 cells in which SPATS2 is very lowly the association of STAU1 to SPATS2. We also observe that expressed. Although we cannot rule out an impact of STAU1 on STAU1 knockdown rescues the apoptotic phenotype resulting SPATS2 translation, our data show that SPATS2 is sensitive to from SNHG5 depletion, thus demonstrating a genetic interaction

10 NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875 ARTICLE between the SNHG5 and STAU1, promoting the stabilization of diluent and incubated in the dark for 10 min at room temperature. Finally, the SPATS2. An intriguing hypothesis is that the up-regulation slides were mounted with DAPI (Life Technologies) nuclear stain. The SNHG5 of SNHG5 during early stages of tumour progression may be probe can be obtained from Exiqon. required in the transforming cells to buffer the STAU1-mediated destabilization occurring on the SPATS2 mRNA, as well as other Vector construction. We amplified the entire SNHG5 genomic locus target mRNAs, hence promoting the survival of the pre-cancerous (±200 bp upstream and downstream the locus) by PCR (HiFi Platinum Taq, 0 cells. Hence, inhibition of STAU1 decay activity on specific Invitrogen) using the following primers: gSNHG5 FW 5 -TACTGGCTGCG- CACTTCG-30, gSNHG5 RV 50-TACCCTGCACAAACCCGAAA-30.The mRNA target provides an intriguing hypothesis for important amplicon was extracted from gel and cloned in pGEM T-Easy vector system early steps in CRC development but additional studies are needed (Promega) and further subcloned into pcDNA 3.1( þ ) (Invitrogen). The to evaluate the overall importance of this mechanism. SNHG5 cDNA sequence was synthetically cloned in pUC57 backbone Altogether, our data describe a role of SNHG5 in regulating cell (Genescript technology) and subcloned into pcDNA 3.1. The pSNHG5 cDNA mutants were obtained using a Quickchange mutagenesis kit (Agilent) following survival in CRC and suggests SNHG5 as a potential candidate for manufacturer’s instructions. The primers used for the mutagenesis are listed in RNA-based anti-cancer drug studies. Supplementary Table 1. The lentiviral vector for the stable expression of SPATS2 cDNA (PLOHS_ccsbBEn_03998, GE Dharmacon) was used as template for subcloning into pcDNA 3.1( þ ) backbone using the following primers: Methods SPATS2 cDNA forward EcoRI 50-GAATTCGTGGCTCTAGAAGGGGAG Cell culture and animal experiments. HCT116, LS174T and HT-29 CRC cell GTGG-30; SPATS2 cDNA reverse EcoRI 50-GAATTCAATTAAATGC- lines were maintained in McCoy’s medium TACTTTCTCATAG-30. (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone) and penicillin/streptomycin (Invitrogen). CACO-2, and SW620 and HEK293 FT cell lines were maintained in DMEM medium (Invitrogen) supplemented with Crystal violet staining. Cells were seeded and reverse transfected in 6-well 10% fetal bovine serum and penicillin/streptomycin. DLD-1 cell line was main- plates. Twenty-four, forty-eight and seventy-two hour after transfection, cells tained in RPMI medium (Invitrogen) supplemented with 10% fetal bovine were washed twice in PBS (Gibco) and fixed with 4% formalin (Sigma) for serum and penicillin/streptomycin. Cell lines were obtained from the ATCC. 1 10 min, and stained with 0.1% crystal violet solution (Sigma). The amount of DNA plasmids were transfected at a final concentration of 2.5 mgml by crystal violet staining was quantified using the ‘ColonyArea’ ImageJ plugin forward transfection using FuGENE HD Transfection Reagent (Promega, Fitch- (Promega). burg, WI, USA) according to manufacturer’s procedures. siRNA oligonucleotides (listed in Supplementary Table 3) were transfected at a final concentration of 50 nM by reverse transfection using RNAiMAX (Invitrogen), according to the EdU incorporation assays. Cells grown in 6-well plates (NUNC) under appro- manufacturer’s instructions. Cells were treated with oxaliplatin (EIMC United priated conditions were pulsed with EdU (33 mM) for 20 min before fixation. The pharmaceutical) in aqueous solution at indicated concentrations and times, EdU was detected using Click-IT EdU Alexa-Fluor 647 Flow Cytometry Assay Kit where appropriate. For the RNA stability assay the cells were treated at the (Molecular Probes, Life Technologies), according to the manufacturer’s protocol. m 1 m indicated time points with 5 gml Actinomycin D or 10 M Triptolide Data were obtained by flow cytometric analysis of 15,000 cells per treatment and (Sigma, St Louis, MO, USA). NMRI-nude adult females were obtained from analysed with the FlowJo 887 software (Tree Star, Ashland, OR, USA). Janvier Labs. All mouse experiments were approved by Dyreforsøgstilsynet according to Danish legislation (licence number 2012-15-2934-00306). Cleaved Caspase-3 staining. Cells were reverse transfected using RNAiMax Tissue samples preparation and microarray analysis. The cancer specimens reagent (Invitrogen) in six-well plates (NUNC). 36 h after transfection the cells included in this study are from the colorectal cancer biobank at the Department were harvested and fixed in 4% formaldehyde methanol-free (Thermo Scientific) of Molecular Medicine, Aarhus University Hospital. A total of 44 MSS and MSI and incubated for 30 min on ice in 90% methanol in 1 PBS pH 7.4 (Gibco). primary stage I–IV (T2-4, N0-3, M0/1) CRCs, 39 adenomas and 20 normal The cells were first stained with anti-cleaved Caspase-3 antibody (#9664, Cell colon mucosa samples were used in the array analyses. Finally, 280 fresh frozen signaling) in 1 PBS-2% BSA (Sigma) for 1 h at room temperature in the dark, microsatellite stable (MSS) or microsatellite instable (MSI), primary stage I-IV washed twice in 1 PBS-2% BSA and stained with Alexa-Fluor 647 goat anti- (T2-4, N0-3, M0/1) CRCs, 33 adenomas and 294 normal colon mucosa were rabbit secondary antibody (Life technology) for one additional hour at room used for validation using RNA-seq. The samples were obtained directly from temperature protected from light. For propidium iodide staining, the labelled cells surgery after removal of the necessary amount of tissue for routine pathology were incubated in propidium iodide (Sigma) and RNase-A (Sigma) solution for examination. Patients who had received preoperative chemotherapy and/or 30 min at 37 °C. Data analysis was performed using the FlowJo 887 software. radiation of rectal cancers were excluded. Postoperatively, the tumours were histologically classified and staged according to the TNM system. Cases with hereditary colorectal cancer syndromes were not included in the study. The use Western blot analysis. Cells were seeded and reverse transfected in 6-well plates of the human tissue samples for research purpose was approved by the Central (NUNC). After 36 h, cells were harvested, washed once with PBS and the pellets Denmark Region Committees on Biomedical Research Ethics (DK; 1999/4678). lysed in RIPA buffer (150 nM NaCl, 0.1% sodium deoxycholate (Sigma), 0.1% SDS Informed written consent was given by all participants. RNA from both tissues (Sigma), 50 mM Tris–HCl (pH 8), 1 mM EDTA (Sigma)) containing protease and cell lines was extracted using the RNeasy Mini kit following the manufacturer’s inhibitors (Complete Mini Protease Inhibitor Cocktail; Roche Applied Science). instructions. The microarray platform and analysis was previously described18. Proteins were separated by electrophoresis in 4–12% NuPAGE Bis–Tris gels Arrays were normalized using the RMA implementation of the oligo software (Invitrogen) and transferred to nitrocellulose membranes. The primary antibodies package and subsequently analysed using the LIMMA package. For differential used were: GAPDH (Santa Cruz, sc-25778), cleaved PARP-1 (Cell signaling, 5625), transcript expression analysis, a linear model was fitted to the expression data. PITRM1 (Pierce, PIEAPA531558), GLE1 (Abcam, ab69968), SPATS2 (Santa Cruz, sc-390306), Bcl-xL (Cell Signaling, 2762), Bcl-2 (Cell Signaling, 2876), STAT3 (Cell Signaling, 9132) and STAU1 (Proteintech, 14225-1-AP). Un-cropped scans of RNA-FISH. HCT116 and TIG-3 human fibroblast were seeded and reverse all western blots presented in the main figures are presented in Supplementary transfected with the indicated siRNAs in chamber glass slides (LAB-TEK). Fig. 8. Twenty-four hour after transfections, the cells were fixed in 3.7% formaldehyde (cat. 28906, Thermo Scientific, Rockford, IL, USA) overnight. Permeabilization was carried out with 70% EtOH at 2–8 °C for at least 60 min followed by Northern blot. Cells were seeded and reverse transfected in 6-well plates (NUNC). 30 min incubation with 0.1 M triethanolamine (Sigma) þ 0.5% (v/v) acetic After 36 h total RNA was isolated using Trizol reagent (Invitrogen). The RNA was anhydride (Sigma) in water at room temperature (RT). The slides were separated by electrophoresis in 10% TBE-Urea gel (Acrylamide 10%, TBE-Urea prehybridized for 1 h at 55 °C with 100 ml of pre-hybridization buffer (50% for- 7M, APS 1 , Temed 1 ) and transferred to nitrocellulose membranes mamide (Sigma) þ 5 SSC þ 5 Denhardts solution þ 0.5% Tween20 (Amersham). The membrane was cross-linked once at 120,000 mJcm 2 with (Thermo scientific)). dFAM-labelled LNA-probes (Exiqon) were added to a final Stratalinker UV Crosslinker (Stratagene). Pre-hybridization was carried out at concentration of 20–30 nM in 100 ml of pre-hybridization buffer and incubated 50 °C for 1 h in Rapid hyb buffer (Amersham). The DNA probes were 32P-labelled at 55 °C for 1 h in a humidified chamber. The slides were washed 3 10 min in with T4 PNK kinase (New England Biolabs) and the hybridization carried out 1.5 h 0.1 SSC (Ambion) at 55 °C, followed by 3 2 min wash in 1 PBS (Gibco) at 50 °C. The membrane were 20 min in 50 ml 5 SSC, 0.1% (w/v) SDS at room at room temperature and subsequently blocked in blocking buffer (10% heat temperature and 2 15 min in 50 ml 1 0.1 SSC, 0.1% (w/v) SDS at 50 °C. The inactivated goat serum (cat. 34864, Sigma) þ 0.5% blocking reagent in PBS-Tween) probe signal was detected using FujiFilm Fluorescent Image Analyzer FLA-3000. for 1 h at RT. For each slide 150 ml anti-FAM-POD diluted 1:40 in blocking DNA probes used were as follows: buffer was added and incubated for 1 h at room temperature. The slides were U50 50-TATCTCAGAAGCCAGATCCG-30; The RNU6B probe was obtained washed twice 2 min in 1 PBS and 50 ml Cy3-TSA substrate diluted 1:50 in from Exiqon.

NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875

Cell fractionation. Cells were grown in 15 cm dishes (NUNC). For each sample EDTA RNAse-free 10 mM, SDS 1%, PMSF 1 , protease inhibitor cocktail, 1 million cells were harvested and the nuclear/cytoplasmic fractionation was Superase-In (Invitrogen)) and sonicated four times (25 cycles each, high pulse performed using Nuclei EZ Lysis Buffer (Sigma), following the manufacturer’s 3000 ON–4500OFF) using a BIORUPTOR (Diagenode). Cell lysates were centrifuged protocol. In the SNHG5 siRNA pool knockdown experiments HCT116 cells were for 20 min @at 4 °C 16,100g and the supernatant aliquoted for the immuno- harvested 36 h after transfection using RNAiMax (Invitrogen). precipitation. The lysates were pre-cleared for 30 min at 4 °C with 50 ml Dyna- beadsMyOne Streptavidin C1 (Life Technologies) in 2 ml of Hybridization solution (NaCl 750 mM, SDS 1%, Tris–Cl pH 7 50 mM, EDTA 1 mM, formamide RNA extraction and qRT–PCR analysis. Total RNA was isolated using Trizol 15%, PMSF 1 , protease inhibitor cocktail, Superase-In, yeast RNA (Ambion) reagent (Invitrogen), treated with TURBO DNase (Ambion, Life Technologies) 0.1 mg ml 1). One ml of hybridization buffer containing 100 pmol of probes and reverse transcribed using TaqMan Reverse Transcription kit (Applied per ml of lysate was added to each sample and incubated for 4 h at 37 °C rotating. Biosystems) using random hexamer primers. QRT–PCRs were performed using In parallel, the beads for the immunoprecipitation were blocked in hybridization Sybr Green PCR Fast PCR Master Mix 2 (Applied Biosystems). The house- solution 30 min at 4 °C. We added 100 ml of beads per 100 pmol of probes keeping genes RPLP0, HPRT1 and GAPDH were used for normalization of and incubated rotating 30 min at 37 °C. The samples were washed ﬁve times qRT–PCR data, unless otherwise stated. The primers used are listed in (5 min each) in 1 ml washing buffer (2 SSC, SDS 0.5%, PMSF 1 )at37°C Supplementary Table 1. rotating. Each sample was Proteinase K (Invitrogen) treated for 45 min at 50 °C shaking before RNA was extracted. A total of 150 ng from each sample were RNA sequencing. HCT116 cells were reverse transfected with All-star negative sent for library preparation and sequencing on the Illumina Hiseq2000 at BGI control (Qiagen), siSNHG5-1 or siSNHG5-2 for 36 h in 6-well plates (NUNC). Genomics (Hong Kong) using Truseq protocol. Primers and biotinylated RNA was extracted from three biological replicates, the quality assessed on the probes sequences used for the pull-down and experiment validation are listed 2100 expert Bioanalyzer (Agilent) and sent for library preparation and in Supplementary Tables 2 and 3, respectively. sequencing on the Illumina Hiseq2000 by BGI Genomics (Hong Kong).

RIA-seq bioinformatics analysis. Trimming. The short reads were trimmed In vitro transcription and copies number quantification. Templates using cutadapt to remove errors such as, for example, trailing adapter sequences. containing the T7 promoter were generated by PCR with high-fidelity DNA The following deviations from default were used: error-rate ¼ 0.05, minimum- polymerase Platinum Taq (Invitrogen). PCR products were purified from agarose length ¼ 25, overlap ¼ 4, quality-cutoff ¼ 20 as well as a list with the relevant gels and 250 ng of template was in vitro transcribed using T7 RNA polymerase Illumina adapter sequences given to the ‘-a’ option. and the MAXIscript kit (Ambion, Life Technologies) for overnight at 37 °C. Mapping. Because the reads were generated from a mixture containing post- After 30 min DNAse treatment with TURBO DNA free at 37 C, the reactions spliced mRNA, it was necessary to use a mapping approach that takes splice were stopped by adding 0.5 mM EDTA. The concentration of the transcript junctions into account. We used the STAR aligner with default options except for was measured by A260 values and converted to the number of copies using outFilterMismatchNoverLmax ¼ 0.1 and outFilterMatchNmin ¼ 16. the molecular weight of the RNA. Dilutions of this transcript were retro-tran- Peak calling. Peak calling was accomplished using the MACS2 algorithm with scribed as described above and the complementary DNA was used as the following non-default options: P ¼ 1e 3, to-large, nomodel and shiftsize ¼ X, standard curve for qPCR. The primers used for qPCR are: SPATS2 T7 where ‘X’ (the peak shift) was estimated individually for each sample using the SPP 0 forward 5 -GGATCCTAATACGACTCACTATAGTAGAAGGGGAG programme. The fLUC samples were pooled and used as reference background. 0 0 GTGGAGGAT-3 ; SPATS2 T7 reverse 5 -GGATCCGCATTGGGCATTTTG The IDR algorithm was used to post-process the MACS2 results and to estimate the 0 0 AAGAA-3 ; SNHG5 T7 forward 5 -GGATCCTAATACGACTCACTATA number of reproducible peaks from the two ODD sample replicates. Owing to a 0 0 GCGGGTGGTAGGAACAATGG-3 ; SNHG5 T7 reverse 5 -GGATCCTTT failed replicate, however, the IDR algorithm could not be applied to the EVEN 0 CATGTTTGTAAAACGAAGAGC-3 . condition, and instead the numbers of reproducible peaks for EVEN were extrapolated from the ODD condition. This gave a conservative result of 736 peaks Luciferase assay. Luciferase reporter constructs containing regions of the SPA and an optimal result of 905 peaks for both conditions, although any peaks called TS2 (NM_001293286.1) and GLE1 (NM_001003722) 30 UTRs were cloned into in intron areas were then subsequently removed, as these are most likely errors pMIR-REPORT firefly luciferase vector (Ambion). pRL-Tk Renilla luciferase arising from an incorrect peak shift. reporter was used as transfection control and luciferase assay normalization. Motif finding. We used the meme-chip programme for motif finding. As input The assays were performed 48 h after transfection of the indicated constructs for meme-chip we used a region of 250 bp flanking each peak summit, for a total of into 2.4 104 HEK293 cells per well (four wells each samples) seeded in 96-well 501 bp long sequences extracted from both ODD and EVEN peak lists. To prevent plates. The cells were transfected with 50 ng of firefly luciferase vectors and 1 ng of biases between the sizes of the two sets, sequences were randomly extracted from the pRL-Tk Renilla reporter. The firefly reporters were competed by co-transfecting the larger set (EVEN) until the size of the smaller (ODD) was reached. Default 50 ng of pcDNA 3.1 ( þ ) expressing the SPATS2 cDNA. The reporter activities parameters were used except that ‘-norc’ was used. This option prevents the were measured using the Dual Glo Luciferase assay System (Promega) and programme from searching the reverse complement, which is irrelevant here as the GloMax Multi Detection System (Promega). The following primers were used sequences under consideration are single-stranded RNA. for the fragment subcloning PmeI–HinIII into pMIR-REPORT: SPATS2 30 UTR forward PmeI primer 50-GTTTAAACGCCTCTATCCCAGAATGTGC-30 and reverse HindIII primer 50-AAGCTTTGCTACTTTCTCATAGACTTCCCTA-30; RNA immunoprecipitation. HCT116 cells were cross-linked using 1500 mJcm 2 GLE1 30 UTR forward PmeI primer 50-GTTTAAACCCTCCCTTGTCTC- and collected in cold PBS. Fifty millions cells per samples were subjected to TAGTGTCTT-30 and reverse Hind III primer 50-AAGCTTCAACCATAACCCA- nuclear/cytoplasmic fractionation using Nuclei EZ Lysis Buffer (Sigma), cyto- GACTTGCAT-30. plasmic fractions were collected and the volume adjusted in 2 polysome buffer (Tris–Cl pH 7.4 20 mM, NaCl 200 mM, MgCl2 2.5 mM, Triton-X 1%, DTT 1 mM, protease inhibitor cocktail, Superase-In). Lysates were pre-cleared with yeast RNA In vivo imaging acquisition to follow tumour development. A luminescent (0.1 mg ml 1) and recombinant protein G agarose (Thermo Scientific) for 30 min variant of the HCT116 cell lines were generated by transduction with the at 4 °C. One per cent of the sample were used as input, the remaining samples pFULT lentiviral vector containing the luciferase2-tdTomato cassette as described 36 incubated with STAU1 antibody (Proteintech, 14225-1-AP) or IgG from rabbit before . The shRNA expressing cassettes harbouring a Tet-inducible promoter ° 37 serum (I5006, Sigma) overnight at 4 C, The RNA-antibody complexes were were generated as described previously by other labs . Two millions freshly collected by incubation with Dynabeads protein G beads (Invitrogen). The samples harvested cells from these shRNA subclones (200 ml from a stock solution of were washed four times and Proteinase K treated before the RNA was extracted for 10 millions cells per ml of PBS) were subcutaneously injected into the right qRT–PCR analysis. flank of 22 weeks old NMRI-nude adult females (Janvier Labs). Tumour growth was measured based on the luminescent signal using the IVIS Lumina II system (Perkin Elmer), every 2–4 days from 1 week after cells injection. Mice were treated with doxycycline (ad libitum, active compound diluted in drinking water at RNA–RNA interaction prediction. Thermodynamic stable interaction sites of 0 2.5 mg ml 1, from Piggidox powder, Boehringer Ingelheim) when shGFP tumours SNHG5 (ENST00000414002), the 3 UTR of SPATS2 (NM 001293285.1) and the 0 reached an average of 100 mm3. The doxycycline solution was changed every 2–3 3 UTR of GLE1 (NM 001003722.1) were predicted with IntaRNA (version 1.2.5; default parameters)38. The interaction sites of lowest energy with SPATS2 days until the end of the experiment. The shRNA sequences are listed in 1 1 Supplementary Table 3. ( 15.50 kcal mol )andGLE1 ( 14.7 kcal mol ) span the exons 1 and 2 of SNHG5. The conservation of the interaction site was tested with PETcofold (version 3.2; maximal percentage of gaps in alignment column -g is set to 0.5)39. RNA interactome analysis and sequencing. HCT116 cells were collected The transcript alignments were extracted from the genome-wide hg19 46way (30 million per sample) and resuspended in 20 ml of PBS containing 1% multiz alignments. We used the ‘Stitch MAF blocks’ function of the Galaxy browser Glutaraldehyde (Sigma) for 10 min at room temperature rotating. Glutaraldehyde to get a connected alignment for each exon, removed all sequences with more than was quenched with 2 ml of 1.25 M Glycine (Sigma) (1/10 total volume) for 50% gaps from the alignments, and concatenated exons of the species that occurred 5 min. The cell pellets were lysed in 2 ml lysis buffer (Tris–Cl pH 7 50 mM, in all exons.

12 NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875 ARTICLE

Statistical testing. All P values are calculated using a two-tailed Student’s 22. Senoo, M., Hoshino, S., Mochida, N., Matsumura, Y. & Habu, S. Identification t-test, unless otherwise stated. For the analysis of the growth curves from the of a novel protein p59(scr), which is expressed at specific stages of mouse xenograft experiment, the following method was applied: Categoric variables spermatogenesis. Biochem. Biophys. Res. Commun. 292, 992–998 (2002). were reported by means of contingency tables. Furthermore, for continuous 23. Park, E. & Maquat, L. E. Staufen-mediated mRNA decay. Wiley Interdiscip. Rev. variables, median and range were computed. To investigate their associations RNA 4, 423–435 (2013). with the product and group parameters, univariate statistical analyses were 24. Lykke-Andersen, S. et al. Human nonsense-mediated RNA decay initiates performed using Kruskal–Wallis or Mann–Whitney non-parametric test. More- widely by endonucleolysis and targets snoRNA host genes. Genes Dev. 28, over, multivariate analyses were carried out. Linear mixed-regression model, 2498–2517 (2014). containing both fixed and random effects, was used to determine the relationship 25. Williams, G. T. & Farzaneh, F. Are snoRNAs and snoRNA host genes new between luminescence growth and number of days after injection. The fixed part players in cancer? Nat. Rev. Cancer 12, 84–88 (2012). of the model included variables corresponding to the number of post-injection 26. Mourtada-Maarabouni, M., Pickard, M. R., Hedge, V. L., Farzaneh, F. & days and the different product treatments. Interaction terms were built into the Williams, G. T. GAS5, a non-protein-coding RNA, controls apoptosis and is model; random intercepts and random slopes were included to take time into downregulated in breast cancer. Oncogene 28, 195–208 (2009). account. The coefficients of the model were estimated by maximum likelihood and considered significant at the 0.05 levels. The power of analysis was reduced 27. Yu, M. et al. High expression of ncRAN, a novel non-coding RNA mapped to because of the limited number of xenograft tumours and few tumours did not chromosome 17q25.1, is associated with poor prognosis in neuroblastoma. Int. have measurements for all variables. All P values reported are two sided and the J. Oncol. 34, 931–938 (2009). significance level was set at 5% (Po0.05) Statistical analysis was performed 28. Montes, M. et al. The lncRNA MIR31HG regulates p16(INK4A) expression to using the STATA 13 software (Stata Corporation, College Station, TX, USA). modulate senescence. Nat. Commun. 6, 6967 (2015). 29. Nakamura, Y. et al. The GAS5 (growth arrest-specific transcript 5) gene fuses to BCL6 as a result of t(1;3)(q25;q27) in a patient with B-cell lymphoma. Cancer Genet. Cytogenet. 182, 144–149 (2008). Data availability. The datasets generated during and/or analysed during the 30. Dong, X. Y. et al. Implication of snoRNA U50 in human breast cancer. J. Genet. current study are available: the CRC microarray data: GSE76713. The CRC Genomics. 36, 447–454 (2009). RNA-seq data: Supplementary Table 4. The RNA-seq data from HCT116 cells 31. Dong, X. Y. et al. SnoRNA U50 is a candidate tumor-suppressor gene at 6q14.3 with knockdown of SNHG5: GSE76801. The RIA-seq data from HCT116 cells for with a mutation associated with clinically significant prostate cancer. Hum. SNHG5: GSE76792. Mol. Genet. 17, 1031–1042 (2008). 32. Siprashvili, Z. et al. The noncoding RNAs SNORD50A and SNORD50B bind K-Ras and are recurrently deleted in human cancer. Nat. Genet. 48, 53–58 (2015). References 33. Alonso, C. R. A complex ’mRNA degradation code’ controls gene expression 1. Torre, L. A. et al. Global cancer statistics. CA Cancer J. Clin. 65, 87–108 during animal development. Trends Genet. 28, 78–88 (2012). ð2012Þ: 34. Wang, J., Gong, C. & Maquat, L. E. Control of myogenesis by rodent SINE- 2. Markowitz, S. D. & Bertagnolli, M. M. Molecular origins of cancer: molecular containing lncRNAs. Genes Dev. 27, 793–804 (2013). basis of colorectal cancer. N. Engl. J. Med. 361, 2449–2460 (2009). 35. Xu, T. P. et al. SP1-induced upregulation of the long noncoding RNA TINCR 3. Reimers, M. S., Zeestraten, E. C., Kuppen, P. J., Liefers, G. J. & van de Velde, C. regulates cell proliferation and apoptosis by affecting KLF2 mRNA stability in J. Biomarkers in precision therapy in colorectal cancer. Gastroenterol. Rep. 1, gastric cancer. Oncogene 34, 5648–5661 (2015). 166–183 (2013). 36. Liu, H. et al. Cancer stem cells from human breast tumors are involved in 4. Djebali, S. et al. Landscape of transcription in human cells. Nature 489, spontaneous metastases in orthotopic mouse models. Proc. Natl Acad. Sci. USA 101–108 (2012). 107, 18115–18120 (2010). 5. FANTOM Consortium. The transcriptional landscape of the mammalian 37. Wiederschain, D. et al. Single-vector inducible lentiviral RNAi system for genome. Science 309, 1559–1563 (2005). oncology target validation. Cell Cycle 8, 498–504 (2009). 6. Tripathi, V. et al. The nuclear-retained noncoding RNA MALAT1 regulates 38. Busch, A., Richter, A. S. & Backofen, R. IntaRNA: efficient prediction of alternative splicing by modulating SR splicing factor phosphorylation. Mol. Cell bacterial sRNA targets incorporating target site accessibility and seed regions. 39, 925–938 (2010). Bioinformatics 24, 2849–2856 (2008). 7. Gupta, R. A. et al. Long non-coding RNA HOTAIR reprograms chromatin 39. Seemann, S. E., Richter, A. S., Gesell, T., Backofen, R. & Gorodkin, J. PETcofold: state to promote cancer metastasis. Nature 464, 1071–1076 (2010). predicting conserved interactions and structures of two multiple alignments of 8. Yap, K. L. et al. Molecular interplay of the noncoding RNA ANRIL and RNA sequences. Bioinformatics 27, 211–219 (2011). methylated histone H3 lysine 27 by polycomb CBX7 in transcriptional silencing of INK4a. Mol. Cell 38, 662–674 (2010). 9. Tseng, Y. Y. et al. PVT1 dependence in cancer with MYC copy-number Acknowledgements 512, increase. Nature 82–86 (2014). We thank Disa E. Tehler and Sophia J. Ha¨fner for commenting on the manuscript; Kim 10. Huarte, M. et al. A large intergenic noncoding RNA induced by p53 mediates Holmstrøm (Bioneer) for help with the RNA in situ hybridization experiments, and 142, global gene repression in the p53 response. Cell 409–419 (2010). Steen Holmberg and Darima Petersen for help with the polysome profile experiments 11. Tay, Y. et al. Coding-independent regulation of the tumourtumor suppressor (Copenhagen University, DK). We thank Pr. Sanjiv Sam Gambhir (Stanford University, 147, PTEN by competing endogenous mRNAs. Cell 344–357 (2011). USA) for providing the pFULT lentiviral vector. Work in the Lund laboratory was 12. Cesana, M. et al. A long noncoding RNA controls muscle differentiation by supported by funding from the People Programme (Marie Curie Actions) of the Eur- 147, functioning as a competing endogenous RNA. Cell 358–369 (2011). opean Union’s Seventh Framework Programme FP7/2007-2013/ under REA Grant 13. Gong, C. & Maquat, L. E. lncRNAs transactivate STAU1-mediated mRNA Agreement 607720, the Danish Council for Independent Research (Sapere Aude pro- 0 470, decay by duplexing with 3 UTRs via Alu elements. Nature 284–288 gramme), the Novo Nordisk Foundation, the Lundbeck Foundation, the Danish National (2011). Research Foundation (DNRF82) and the Danish Cancer Society. 14. Heraud-Farlow, J. E. & Kiebler, M. A. The multifunctional Staufen proteins: conserved roles from neurogenesis to synaptic plasticity. Trends Neurosci. 37, 470–479 (2014). 15. Ricci, E. P. et al. Staufen1 senses overall transcript secondary structure to Author contributions regulate translation. Nat. Struct. Mol. Biol. 21, 26–35 (2014). N.D.D. and A.H.L. designed the experiments, analysed the data and wrote the manu- 16. Kim, Y. K. et al. Staufen1 regulates diverse classes of mammalian transcripts. script. All coauthors commented on the manuscript. N.D.D., M.M., C.C. and H.M.G. EMBO J. 26, 2670–2681 (2007). conducted the experiments. M.M.N., C.F.R., S.S., N.R., S.V. S.T. and J.S.P. performed 17. Sugimoto, Y. et al. hiCLIP reveals the in vivo atlas of mRNA secondary bioinformatics analyses. L.L.C., T.Ø. and C.L.A. contributed clinical samples and analyses structures recognized by Staufen 1. Nature 519, 491–494 (2015). hereof, and R.B. contributed technical expertise on the RIA-seq protocol. All authors 18. Nielsen, M. M. et al. Identification of expressed and conserved human commented on the manuscript. noncoding RNAs. RNA 20, 236–251 (2014). 19. Derrien, T. et al. The GENCODE v7 catalog of human long noncoding RNAs: analysis of their gene structure, evolution, and expression. Genome Res. 22, 1775–1789 (2012). Additional information 20. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based Supplementary Information accompanies this paper at http://www.nature.com/ approach for interpreting genome-wide expression profiles. Proc. Natl Acad. naturecommunications Sci. USA 102, 15545–15550 (2005). 21. Kretz, M. et al. Control of somatic tissue differentiation by the long non-coding Competing financial interests: The authors declare no competing financial RNA TINCR. Nature 493, 231–235 (2013). interests.

NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13875

Reprints and permission information is available online at http://npg.nature.com/ This work is licensed under a Creative Commons Attribution 4.0 reprintsandpermissions/ International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise How to cite this article: Damas, N. D. et al. SNHG5 promotes colorectal cancer cell in the credit line; if the material is not included under the Creative Commons license, survival by counteracting STAU1-mediated mRNA destabilization. Nat. Commun. users will need to obtain permission from the license holder to reproduce the material. 7, 13875 doi: 10.1038/ncomms13875 (2016). To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Publishers’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afﬁliations. r The Author(s) 2016

14 NATURE COMMUNICATIONS | 7:13875 | DOI: 10.1038/ncomms13875 | www.nature.com/naturecommunications